A systematic literature review: Recent techniques of predicting STEM stream students

Environmental Factors: Environmental factors category looks at how support from friends and families, academic environments, including mentoring relationships, and residential environments would have a more positive outcome and expectations in the STEM field, and they serve as a protective factor in academically challenging environments, as well as an important socialising agent in determining the field of study choice. The variable of parental encouragement was discovered to be a critical developmental predictor that laid the groundwork for a STEM career (Wang and Zhao, 2012; Wang, Ye and Degol, 2017). The breadth of teacher influence variables include anything from assisting students in improving their academic success and self-assurance to maximising student exposure to mathematics and science courses, which was a significant predictor of STEM growth (Heaverlo, 2011; Sellami, El-Kassem, Al-Qassass and Al-Rakeb, 2017; Wang and Kenny, 2011). Teachers can take various steps in the classroom to improve student engagement and confidence (Rabenberg, 2013). Student's STEM career aspiration variable has a remarkable influence on the environmental factor category in students’ STEM prediction (Radzi and Sulaiman, 2018). Researches had shown that students who have a clear and defined career goal in a STEM field were more likely to persist in their studies and eventually enter into a STEM profession. One of the main factors contributing to their career aspiration in STEM was the recruitment opportunities after degree. Some literatures also considered financial aid such as scholarship and merit-based aid program affected students’ college decisions in STEM fields major (Sjoquist and Winters, 2015; Sinatra, Brem, & Evans, 2008). Financial aid made available to students in urban areas (Ansong, Okumu, Albritton, Bahnuk, & Small, 2020), may make STEM education more accessible than it is in rural regions. Financial hurdles were cited as a reason why high school students chose not to pursue STEM areas in college. Indeed, Wang and Kenny (2011) found a link between financial aid and being driven to do well in mathematics and pursuing STEM careers in college.The findings of this research highlighted the enormous potential of the ML-based prediction method, the newest development in STEM stream predictions, as well as the need for more research in the future. Although only a few articles reported about the use of ML, it is showing an upward trend in recent publications. There is also the fact that ML is open to potential improvement that has yet to be considered. There is evidence that ML techniques have become increasingly popular for prediction and many decisions have been made as a result of the effective and efficient use of data mining. As an example, in an educational dataset, ML techniques can be used to uncover hidden patterns and create a prediction system that can forecast a group of secondary school subjects (e.g. Science, Business Studies, and Humanities) (Hasan et al., 2020). The classification technique in supervised ML was found to be more popular in the review, with 11 articles, compared to Deep Learning-based approaches, which had only two. According to the literatures, the classification method, predicts a discrete output value that can be implemented utilising a number of different algorithms: Support Vector Machine (SVM), including Naive Bayes (NB), Logical Regression (LR), Random Forest (RF), Gradient Boosting Decision Tree (GBDT), k-Nearest Neighbour (kNN), Decision Tree (DT), J48, and Bayesian Model algorithm. However, none of the papers mentioned that a regression approach such as Linear Regression or Decision Trees Regressor, which predicts a continuous value output, may be used. Despite the fact that unsupervised learning techniques were rarely discussed in the literature, it is interesting to note that numerous clustering algorithms, such as Fuzzy and K-means, have been applied (2 articles). This strategy is advantageous when the data set has no labels applied to it. It looks for tendencies that have not been seen before but with the least amount of human examination.The authors would like to express their gratitude to Universiti Sains Malaysia (USM) and Universiti Malaysia Perlis (UniMAP) for their support in completing this study.

Environmental Factors: Environmental factors category looks at how support from friends and families, academic environments, including mentoring relationships, and residential environments would have a more positive outcome and expectations in the STEM field, and they serve as a protective factor in academically challenging environments, as well as an important socialising agent in determining the field of study choice. The variable of parental encouragement was discovered to be a critical developmental predictor that laid the groundwork for a STEM career (Wang and Zhao, 2012; Wang, Ye and Degol, 2017). The breadth of teacher influence variables include anything from assisting students in improving their academic success and self-assurance to maximising student exposure to mathematics and science courses, which was a significant predictor of STEM growth (Heaverlo, 2011; Sellami, El-Kassem, Al-Qassass and Al-Rakeb, 2017; Wang and Kenny, 2011). Teachers can take various steps in the classroom to improve student engagement and confidence (Rabenberg, 2013). Student's STEM career aspiration variable has a remarkable influence on the environmental factor category in students’ STEM prediction (Radzi and Sulaiman, 2018). Researches had shown that students who have a clear and defined career goal in a STEM field were more likely to persist in their studies and eventually enter into a STEM profession. One of the main factors contributing to their career aspiration in STEM was the recruitment opportunities after degree. Some literatures also considered financial aid such as scholarship and merit-based aid program affected students’ college decisions in STEM fields major (Sjoquist and Winters, 2015; Sinatra, Brem, & Evans, 2008). Financial aid made available to students in urban areas (Ansong, Okumu, Albritton, Bahnuk, & Small, 2020), may make STEM education more accessible than it is in rural regions. Financial hurdles were cited as a reason why high school students chose not to pursue STEM areas in college. Indeed, Wang and Kenny (2011) found a link between financial aid and being driven to do well in mathematics and pursuing STEM careers in college.The findings of this research highlighted the enormous potential of the ML-based prediction method, the newest development in STEM stream predictions, as well as the need for more research in the future. Although only a few articles reported about the use of ML, it is showing an upward trend in recent publications. There is also the fact that ML is open to potential improvement that has yet to be considered. There is evidence that ML techniques have become increasingly popular for prediction and many decisions have been made as a result of the effective and efficient use of data mining. As an example, in an educational dataset, ML techniques can be used to uncover hidden patterns and create a prediction system that can forecast a group of secondary school subjects (e.g. Science, Business Studies, and Humanities) (Hasan et al., 2020). The classification technique in supervised ML was found to be more popular in the review, with 11 articles, compared to Deep Learning-based approaches, which had only two. According to the literatures, the classification method, predicts a discrete output value that can be implemented utilising a number of different algorithms: Support Vector Machine (SVM), including Naive Bayes (NB), Logical Regression (LR), Random Forest (RF), Gradient Boosting Decision Tree (GBDT), k-Nearest Neighbour (kNN), Decision Tree (DT), J48, and Bayesian Model algorithm. However, none of the papers mentioned that a regression approach such as Linear Regression or Decision Trees Regressor, which predicts a continuous value output, may be used. Despite the fact that unsupervised learning techniques were rarely discussed in the literature, it is interesting to note that numerous clustering algorithms, such as Fuzzy and K-means, have been applied (2 articles). This strategy is advantageous when the data set has no labels applied to it. It looks for tendencies that have not been seen before but with the least amount of human examination.The authors would like to express their gratitude to Universiti Sains Malaysia (USM) and Universiti Malaysia Perlis (UniMAP) for their support in completing this study.

Environmental Factors: Environmental factors category looks at how support from friends and families, academic environments, including mentoring relationships, and residential environments would have a more positive outcome and expectations in the STEM field, and they serve as a protective factor in academically challenging environments, as well as an important socialising agent in determining the field of study choice. The variable of parental encouragement was discovered to be a critical developmental predictor that laid the groundwork for a STEM career (Wang and Zhao, 2012; Wang, Ye and Degol, 2017). The breadth of teacher influence variables include anything from assisting students in improving their academic success and self-assurance to maximising student exposure to mathematics and science courses, which was a significant predictor of STEM growth (Heaverlo, 2011; Sellami, El-Kassem, Al-Qassass and Al-Rakeb, 2017; Wang and Kenny, 2011). Teachers can take various steps in the classroom to improve student engagement and confidence (Rabenberg, 2013). Student's STEM career aspiration variable has a remarkable influence on the environmental factor category in students’ STEM prediction (Radzi and Sulaiman, 2018). Researches had shown that students who have a clear and defined career goal in a STEM field were more likely to persist in their studies and eventually enter into a STEM profession. One of the main factors contributing to their career aspiration in STEM was the recruitment opportunities after degree. Some literatures also considered financial aid such as scholarship and merit-based aid program affected students’ college decisions in STEM fields major (Sjoquist and Winters, 2015; Sinatra, Brem, & Evans, 2008). Financial aid made available to students in urban areas (Ansong, Okumu, Albritton, Bahnuk, & Small, 2020), may make STEM education more accessible than it is in rural regions. Financial hurdles were cited as a reason why high school students chose not to pursue STEM areas in college. Indeed, Wang and Kenny (2011) found a link between financial aid and being driven to do well in mathematics and pursuing STEM careers in college.The findings of this research highlighted the enormous potential of the ML-based prediction method, the newest development in STEM stream predictions, as well as the need for more research in the future. Although only a few articles reported about the use of ML, it is showing an upward trend in recent publications. There is also the fact that ML is open to potential improvement that has yet to be considered. There is evidence that ML techniques have become increasingly popular for prediction and many decisions have been made as a result of the effective and efficient use of data mining. As an example, in an educational dataset, ML techniques can be used to uncover hidden patterns and create a prediction system that can forecast a group of secondary school subjects (e.g. Science, Business Studies, and Humanities) (Hasan et al., 2020). The classification technique in supervised ML was found to be more popular in the review, with 11 articles, compared to Deep Learning-based approaches, which had only two. According to the literatures, the classification method, predicts a discrete output value that can be implemented utilising a number of different algorithms: Support Vector Machine (SVM), including Naive Bayes (NB), Logical Regression (LR), Random Forest (RF), Gradient Boosting Decision Tree (GBDT), k-Nearest Neighbour (kNN), Decision Tree (DT), J48, and Bayesian Model algorithm. However, none of the papers mentioned that a regression approach such as Linear Regression or Decision Trees Regressor, which predicts a continuous value output, may be used. Despite the fact that unsupervised learning techniques were rarely discussed in the literature, it is interesting to note that numerous clustering algorithms, such as Fuzzy and K-means, have been applied (2 articles). This strategy is advantageous when the data set has no labels applied to it. It looks for tendencies that have not been seen before but with the least amount of human examination.The authors would like to express their gratitude to Universiti Sains Malaysia (USM) and Universiti Malaysia Perlis (UniMAP) for their support in completing this study.